Most articles on this topic come from vendors who sell something other than blanket discounts. The deep-discount study that turns up in this debate was not done in ecommerce. We separate what is evidenced from what is a rule of thumb, then show the arithmetic you can run yourself. This piece sits inside a wider chain: ROAS looks good, profit doesn't.
What does the research say, and where was it done?
Less than the confident headlines imply.
A university summary describes a Journal of Retailing study on whether deep promotional discounts work. It used transactions from a US grocery chain's physical stores, not an online shop1. That matters. Shipping, returns, paid acquisition and price comparison all work differently online. Treat it as a related finding, not a verdict on your site.
Academic work on the long-term effects of price promotions also exists. A secondary summary of a Journal of Marketing Research paper covers it, using weekly scanner-panel sales for one perishable and one storable product2. Again: a different category, a different sales channel.
We did not find clean, independent, ecommerce-specific evidence that settles the question. That is a finding too. It means any single number you read about "the effect of discounting" deserves a raised eyebrow.
A discount takes a bigger share of margin than of price.
What happens to margin when you discount?
Start with contribution margin. In our convention, that is revenue minus the variable costs of an order, before marketing: cost of goods, shipping, payment fees, returns and discounts.
A vendor post on this topic makes the same basic point: a discount hits margin harder than it hits revenue3. You do not need the vendor to see why. The discount comes off the price, but the costs of the order stay where they were.
Example (illustrative): A product sells for €100. Its other variable costs are €60, so contribution margin is €40 per order. Take 20% off. The price is €80, the costs are still €60, and margin falls to €20. The customer saw a one-fifth price cut. You lost half your margin.
The extra volume you need to stand still follows from that. Take margin (m) and discount (d), both as a share of the original price.
Orders needed to break even = m ÷ (m − d).
Example (illustrative): With a margin of 40 per 100 of price, here is what each discount level asks of your volume.
| Discount (d) | Margin per 100 of original price | Orders needed to match the old margin |
|---|---|---|
| None | 40 | The same as now |
| 10 points | 30 | A third more |
| 20 points | 20 | Double |
| 30 points | 10 | Four times as many |
The table is a model, not a benchmark. Your margin will differ. The shape will not.
There is a second leak. Not every order in a promotion is incremental. Some buyers would have purchased anyway, at full price. That is your baseline: the sales you would expect without the activity. Those buyers now get the discount for free.
Example (illustrative): Without a promotion you would sell 1,000 orders at €40 margin: €40,000. With a 20% discount you sell 1,300 orders at €20 margin: €26,000. Revenue rose. Margin fell by €14,000. To break even you would need 2,000 orders.
Add the discount rate you apply to the whole basket, stacked codes and free shipping, and the hole gets deeper. Revenue is lovely. Keeping some of the money is also worth a meeting.
Do customers acquired with a discount come back?
This is where discounts could still win: a first order at low margin, then repeat orders at full price. It is also where the evidence is weakest.
The question is about retention and customer lifetime value (LTV). Figures on how discount-acquired customers behave do circulate, mostly from vendors. We could not trace one to data we were able to check, so we do not repeat them here.
The honest position is: nobody outside your business knows. Some discount-acquired customers return and some never do. The mix depends on your product, your price and how often you run sales. Which brings us to the only evidence that counts: yours.
Black Friday: what the evidence shows and what it doesn't
Black Friday is where the pressure to discount is highest, so it is where the claims are loudest.
What we can say: we found no independent, checkable ecommerce data on how Black Friday customers behave afterwards. What we cannot say: that they are worth more, or less, than your other customers. Your own cohorts can answer that. A headline about someone else's customer base cannot.
What is worth doing is separating three things in your own data:
- Orders you would have had anyway. Your baseline for that week.
- Orders pulled forward. Sales from December or January that moved into November.
- Orders that are new, and come back. The only group that justifies a deep cut.
If your sales dip in the weeks after a promotion, some of what looked like growth was borrowed.
How do you test a promotion on your own data?
You have three practical tools. Use them in this order.
- Compute break-even uplift before you launch. Use the formula above with your real contribution margin. If the uplift you need looks implausible, stop there.
- Compare against a baseline. Write down what you expect without the promotion before it starts. A baseline chosen afterwards is a story. If you can hold some customers or regions back, a holdout test gives a cleaner answer. This is the idea behind incrementality: what happened because of the promotion, over and above what would have happened anyway.
- Follow the buyers. Use cohort analysis: group customers by the month and the offer of their first order, then see how they behave over the following months. Compare discount-acquired cohorts with full-price ones, and be careful about how different the two groups were to begin with.
Judge the promotion on incremental margin and on repeat purchase, not on revenue. If you want to put your own costs into the maths, the Break-even ROAS & contribution margin calculator is a start. For the reported-versus-real ROAS question, try the Is your ROAS real? calculator.
When do discounts make sense?
Not never. A discount is a tool, and some uses are defensible:
- Clearing stock that costs you cash to hold, where the margin you lose is smaller than the cash you free up.
- A planned entry offer where you have cohort data showing those customers come back.
- A deliberate price test with a baseline and a stop rule.
What is hard to defend is a standing discount code that has an active social life on coupon sites, applied to customers who were about to buy. That one is not marketing. It is a price cut with extra steps.
For ZFD: We judge a promotion on contribution margin and repeat behaviour, not on the revenue line, and we say plainly when the data cannot settle it.